|Publication number||US20020035573 A1|
|Application number||US 09/952,827|
|Publication date||Mar 21, 2002|
|Filing date||Sep 14, 2001|
|Priority date||Aug 1, 2000|
|Also published as||US7464086|
|Publication number||09952827, 952827, US 2002/0035573 A1, US 2002/035573 A1, US 20020035573 A1, US 20020035573A1, US 2002035573 A1, US 2002035573A1, US-A1-20020035573, US-A1-2002035573, US2002/0035573A1, US2002/035573A1, US20020035573 A1, US20020035573A1, US2002035573 A1, US2002035573A1|
|Inventors||Peter Black, Anthony Waters|
|Original Assignee||Black Peter M., Waters Anthony Bryan|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (5), Referenced by (43), Classifications (7), Legal Events (4)|
|External Links: USPTO, USPTO Assignment, Espacenet|
 This application is a continuation-in-part of our application No. Ser. No. 09/630,227 filed Aug. 1, 2000 and our application Ser. No. 09/703,006 filed Oct. 31, 2000 and our application Ser. No. 09/738,471 filed Dec. 15, 2000.
 The invention relates to providing a system for sensing and responding to consumer behavior as evidenced by the consumer's selection and consumption of content, products or services over open or closed digital networks, in forms ranging from simple text to complex multimedia.
 The Problem
 Open network systems like the Internet and closed network systems such as those operated by cable television and telephone companies deliver trillions of words and millions of hours of digitized audio and video to billions of computer and television screens. Systems exist which survey traffic on these systems to determine the behavior of consumers. Some systems exist which will identify consumer behavior on the basis of the selection of a particular web page or a particular television program. No system exists however, to analyze and/or survey statistics revealing the underlying interests (psychographic or psycholinguistic behavior) of those persons selecting particular content and portions of that content, to recommend related products, services and content that can be consumed or purchased by the consumer. It would be highly beneficial to create markets on a near real-time basis for those products and services of interest to persons that are already recognized to be interested in a particular related subject.
 Origins of the Solution
 During and immediately following World War II, large scale computing was first applied to the task of managing the explosion of information. Vannevar Bush, FDR's technology czar, laid out the problem in an article in the Atlantic Monthly called ‘As We May Think’ (see http://www.theatlantic.com/unbound/flashbks/computer/bushf.htm ), and imagined a solution—called the MEMEX—which was the precursor to the massively indexed databases and search engines in wide proliferation today. At roughly the same time, Claude Shannon of MIT and Bell Labs (Bush and Shannon knew each other and worked together in the design and deployment of the first computers) laid out ‘Information Theory’ (see http://www.nvu.edu/pages/linguistics/courses/v610003/sha n.html), and the conceptual framework for digital noise reduction, based on the fundamental precepts of Boolean logic.
 Though cloaked in secrecy for decades, the National Security Agency (NSA) has made extensive use of massive scale computing to perform traffic analysis on electronic/digital communications (telephony, telegraphy, RTTY, fax, email, etc.). The standard methodologies employ two different but complementary approaches, forecast by Bush and Shannon: filtering based on Boolean search techniques, and word frequency analysis. The first methodology takes impossibly large arrays of data and produces manageable subsets relevant to the search criteria (‘associative trails’ as imagined with Bush's MEMEX: “Wholly new forms of encyclopedias will appear, ready made with a mesh of associative trails running through them”), the second methodology identifies pervasive themes and/or subject matter within these manageable subsets (in effect, road maps). The resulting analysis can then be ‘fed back’ (feedback is a key concept in Information Theory) into the search process in order to refine and more precisely target the searches.
 Massive computing and associated databasing began to impact the internal operations of big business and the military in the 1950's, somewhat lagging behind the intelligence agencies. In the 1960's, massive computing enabled large scale electronic transaction processing and billing, with consumers benefiting through the arrival of credit cards. For business, the resulting transaction databases enabled datamining for customer behavior profiles, and led to consumer targeting through direct mail and telemarketing. Using settop boxes and diaries, Nielsen and other firms sought to sample consumer behaviors, and used computer-driven statistical analysis and inference to characterize consumer behavioral trends.
 In the early 1980's massive computing became sufficiently inexpensive for academics to employ. Then, the first word frequency analysis projects were undertaken on very large samples of published English language prose, and by the late 1980's the results were commonly available in public literature.
 In the early 1990's, the Office of Naval Research (ONR) embellished word frequency analysis techniques in order to automate the review of international science and technology literature, to create comprehensive conceptual roadmaps through the material.
 The idea was to use machine analysis to figure out what the Russians, and other adversaries and allies were doing in science and technology by using computational linguistics on a closed system of published literature. The result is a technology called Database Tomography (DT), which automates
 the retrieval of relevant documents
 the identification of technical infrastructure (who is citing who, etc.)
 the identification of technical themes and relationships
 the discovery of non-obvious underlying themes in the literature
 In the mid-nineties a further embellishment of word frequency analysis evolved in the academic/technology community, called latent semantic indexing (LSI). LSI seeks to identify the underlying concepts in documents, and then draw conclusions with regards the similarity/relevance to other documents by comparison of the documents thematic matrices.
 In the late 1990's, largely in response to the demands for improved search and ad targeting over the Internet, a number of search enhancement and content analysis techniques were in development.
 Some of these systems required manual intervention. In one instance, Yahoo employed a large numbers of ontologists to develop a knowledge classification system with upwards of 30,000 nodes, in order to assist the search for related material. In another, a firm called Gotuit developed systems for adding additional data (metadata) to streaming audio and video that allowed the material to be ‘sliced and diced’, thus enabling search for specific segments.
 Some of these systems were automatic. In one instance, Rulespace sought to duplicate Yahoo's ontological approach in an automated fashion. Autonomy, and other like firms, sought to automatically classify content according to extent advertising categories. Predictive Networks, and other like firms, sought to classify consumer behavior patterns by tracking consumer's use of clicks and keystrokes while using the Internet.
 The system of this invention, (called the Etronica system) directly tracks what consumers are interested in, by sensing their search behavior.
 Component Methodologies of the Etronica System
 Word frequency analysis on large corpi of English language prose to identify a base keyword set.
 Word frequency analysis on smaller ‘special’ corpi of English language prose (eg. An Electronic Program Guide used in a cable television system, or a law citation database) in order to identify statistically frequent, and hence special ‘terms of art’ for inclusion as extensions to the base keyword set.
 Automated assignment (metatagging) of keywords, drawn from a master keyword set, to individual documents, or records within a database.
 Exploitation of the ‘tagged’ keywords to form effective Boolean ANDed searches.
 Exploitation of the ‘tagged’ keywords as indicators of consumer's territories of interest.
 Signaling consumer interests over a network for centralized accumulation in a datamining system for traffic analysis.
 Exploitation of statistically significant consumer patterns of interest for optimization of ad and merchandise sales and delivery of relevant content.
 For example, while searching the Internet for an article on basketball, various basketball-related television programs or video-on-demand (pay-per-view) movies could be recommended, as well as various products that could be suggested on the screen for purchase, such as sports supplies, sports clothing and books and magazines on the subject of basketball. If it could be determined that the searcher was particularly interested in professional basketball, the products suggested could be narrowed to be more relevant to that interest. Alternatively., while watching a broadcast television program like WEST WING, various related politically-oriented television broadcasts, in dramatic, news and documentary genres (for example a documentary on the Secret Service), could be recommended, as well as related Pay Per View motion pictures (for example, a film such as In The Line Of Fire, through Video on Demand services), as well as an array of related products and services, and related websites, might be recommended. Further, psychographically related products and services, related by coincident behavior rather than common themes of interest, might be incorporated into the recommendations.
 Superiorities of the Invention (the Etronica System)
 1) It is founded on a broad model of human interests and activities, as empirically indicated by the keyword set derived from word frequency analysis of massive, non-specialized corpi of English language prose. The document-specific analysis of LSI and DT limits the reach of the analysis to the system of documents reviewed, and suffers from increasing complexity as documents are added to the system. The advertising specific approach of Autonomy and others limits the analysis to a crude breakdown of advertising categories.
 2) Unlike the numbers-based LSI (and other Neural Net systems), the Etronica system uses a set of tokens based on keywords whose meaning is clear, and easily understood and interpreted by humans.
 3) Unlike computationally intensive systems like LSI (and other Neural Net systems), the Etronica system is fast, and computationally highly efficient. The creation of the keyword set is already done, and the keyword matching to content is principally based on table lookup techniques. The computational requirements grow in a flat, symmetric fashion with the number and length of the documents or records, rather than exponentially, as with LSI, and other matrix-analysis based systems.
 4) Because virtually all digitally searchable bodies of content can be manipulated using Boolean search operators (AND, OR, NOT), the exploitation of the metatagged keywords in the Etronica system to form Boolean ANDed queries is naturally compatible with the de facto international API (application program interface) for search.
 5) Because the Etronica system is founded on an empirically valid keyword set (see 1), tuning the keyword set to a new specialized corpus simply requires the identification of an extension to the base set of keywords, rather than the complete reformation of the set (as is required by most metatagging systems). This is quickly and easily accomplished by a word frequency analysis on the specialized corpus, and comparison of the results to the existing Etronica keyword set to determine the significant differences. This process is, in essence, a feedback loop for signal correction.
 6) Because the Etronica system tracks consumer interest, rather than their transactions (as in the case of Amazon's metatagging system, and many advertising-driven systems), no invasion of individual privacy as a result of the association of individual information with sensing data is either necessary or inevitable in the datamining/traffic analysis process.
 7) Because the Etronica system exploits only most commonly used words in the keyword set as second operands in Boolean ANDed queries and analysis, the synonymy problem suffered by most computation linguistics problems (including DT and LSI) is attenuated.
 8) Because the Etronica system exploits keywords with non-ambiguous meanings (‘movie’, as opposed to ‘film’), the polysemy problem suffered by most computational linguistics systems (including DT and LSI) is attenuated.
 9) Because the Etronica system is based on constant Traffic Analysis, rather than sampling and statistical inference (as practiced by Nielsen, Mediametrix and other consumer sensing systems), and senses human interests, rather than mouseclicks and keystrokes, the resulting profiling of behavior is far more accurate.
 10) Because the Etronica system exploits a ‘flat’ set of keywords (where no words holds a parent-child hierarchical relationship to another, nor is any specific value-based weighting give one keyword over another), rather than the hierarchical systems employed by Yahoo and Rulespace, and derived by DT and LSI, the statistical occurrence of Etronica keywords can be viewed in a combinatorial fashion. In effect, two or more keywords co-occurring in a statistically significant fashion will describe a territory of consumer interest in a more precise fashion, because they have been Boolean ANDed together.
 11) Because the distribution of the Etronica keywords is consistent, and the set of keywords is limited, the storage and transmission of consumer behavior data equipped with a payload of Etronica keywords requires a very small amount of data to be transferred, unlike most other consumer remote-sensing techniques.
 It is an object of this invention to determine the interests of users of closed and open networks by enhancing the content they will select and consume through metatagging, sensing their behavior as they navigate through complex systems of metatagged information and content, recording (either locally on a mass-storage device, or remotely in an aggregated datamining operation) the raw evidence of their behavior and then performing traffic analysis on the raw data to determine statistically significant patterns of behavior which can be used as the basis for 1) improving and optimizing graphic user interfaces (GUI) on computer, television and other kinds of screen displays, 2) making recommendations for related content, products and services and 3) enabling value-based sales of advertising, sponsorships and merchandising opportunities based on traffic analysis-based empirical evidence as to user interests and behavior.
 The METATAG-BASED DATAMINING of this invention uses a manual or automated keyword tagging system to determine content, products and services that share common themes of interest for users. As the directory which references content, products and services is being prepared for publishing on a network, each record is tagged with one or more pre-selected keywords either manually or using an automated tagging process as described in our parent patent applications on “Hotwording” (Ser. No. 09/630,227), “Brilliant Queries” (Ser. No. 09/703,006) and “Targeted E-commerce” (Ser. No. 09/738,471). These keywords are attached (metatagged) to individual records in reference directories of television and cable program information (Electronic Program Guides or EPGs), film directories (used for Video-on-Demand services), and any and all reference directories that might include news, weather, sports, financial and other kinds of special or encyclopedic reference information.
 As users navigate through the metatagged content from the directories, small efficient records of their navigation behavior (enhanced with metatagged keywords that reveal what the user is interested in) are sent for storage in databases (either locally on the PC, mobile device, set-top box or other form of connected appliance, or in a central mass storage repository for datamining (often referred to as a datacube), or both.
 Different classes of user navigation events can be recorded. For example, the simple selection of a metatagged news article would produce a traffic analysis record exposing the user's selection of a specific piece of content, and the ‘angles of interest’ as betrayed by the metatagged keywords. The title or subject of the article, when combined with the metatagged keywords, allows for efficient contextualization.
 Contextualization through tagging of content occurs because, whether explicit or implicit, all content must have a context to allow for clear understanding. For example, an article titled Blackbird might be about a bird, or a song by the Beatles, or a sophisticated reconnaissance airplane. The ‘metatagging’ of the keyword ‘reconnaissance’ to the article in question would leave little or no ambiguity as to the subject matter.
 Since the goal of this technology is to locate specific and well-targeted communities of users who share common interests and behaviors, it is important to note that the analysis of the traffic in a given network must find the logical intersection of all content, product or service selections that match the context of the user's original selection and all possible content, product or service selections that are a probable match for the user's interests and behavior.
 In another example, the use of a ‘Hot Word’ as described in our parent patent application, within a metatagged article (see our patent application for “Hotwording Query System”), would produce a traffic analysis record detailing the user's impulsive selection of a subject of interest (effectively, the first operand in a Boolean ANDed specification of the user's territory of interest) associated with the keywords metatagged to the article (effectively, an array of second operands in a Boolean ANDed specification of the user's territory of interest).
 In another example, the use of a “Brilliant Query” as described in our parent patent application, associated with a metatagged article (see patent application for “Brilliant Query”), would produce a traffic analysis record detailing the user's analytically driven selection of a recommended search (effectively a Boolean NOT specification of the user's territory of interest by virtue of the fact of the exclusion of the other recommended queries, each of which exploits one of the metatagged keywords as a second operand in a Boolean ANDed search, and thus represents a specific angle of interest).
 In another example, the use of a “Targeted E-Commerce” offer as described in our parent patent application, associated with a metatagged article (see patent application for “Targeted E-Commerce”), would produce a traffic analysis record detailing the user's selection of a recommended product or service (effectively, the product or service serving as the first operand in a Boolean ANDed specification of the user's territory of interest) associated with the keywords metatagged to the article and shared in the product or service record in the directory (effectively, an array of second operands in a Boolean ANDed specification of the user's territory of interest), and betraying the user's behavior and interests when he or she is prepared to consider an economic transaction.
 Content, products or services are associated with one or more keywords, which are manually or automatically selected. An example of keywords associated with, for instance, “Basketball”, would be “clothing” and/or “celebrity”. An example of keywords associated with, for instance, an episode of the television show “WEST WING”, would be “politics” and/or “president” and/or “assassination”. All example of keywords associated with, for instance, a news article on the SR-71 ‘Blackbird’ airplane, would be “aviation” and/or “reconnaissance”.
 All keywords are listed in a keyword table. Recommendations are made by performing a statistical frequency analysis of keyword occurrence in the datamined records of user navigation events (either differentiated by class of event, or taken in aggregate) and then comparison to the metatagged keyword ‘signatures’ of individual articles of content in the directory of content, products and services to determine direct relevance (a precise match), or close relevance (a close match when multiple keywords are considered). As discussed in our “Brilliant Query” application, more definitive results will be obtained if keyword clusters, such as a doublet (a two word combination) or a triplet (tliree word combination) are analyzed, as they are far more revealing than a single word or general content category.
 The recommendation of content, products or services, is a simple process of filtering the total array of alternatives in order to limit the display to only those associated with specific keywords or multiple keyword clusters. The filtering is accomplished through conventional Boolean AND and NOT operations on a database of user navigation records, where every record is fortified by metatagged keywords.
 Special ‘pilot’ keywords can be added to select records, in order to flag special demographic or psychographic characteristics in the process of datamining. For example, a program guide entry for the television program “WEST WING” might be tagged with a special ‘pilot’ keyword to identify “WEST WING” watchers, which will reflect the unique demographics of the people who watch “WEST WING”. Datamining would reveal what other content, products and services people who watch “West Wing” are inclined to be interested in, based on statistical analysis of coincident behavior. Marketers generally refer to these kinds of content, products and behavior as ‘adjacent’.
 Selection for recommendation can be done statically as a preprocess on reference directories, thus enabling the caching of recommendations on either the host or the local device or the selection for recommendation can be performed on the host or local device dynamically.
 With the static publishing model, the selection for recommendation is done at the time the products, content or services are submitted for publishing (normally in a closed network, where the network operator has full control over what is going to be published) in the directory and the recommendations are fully integrated with the hosted reference directory. This model serves the passive user.
 The dynamic publishing model relies on a reference directory that is running live on the server and gateway interface technology to query the reference directory and produce recommendations as the user navigates through each new alternative selected from the total array of content, products and services. This model allows for recommendations that are instantly updated as the user actively searches for new alternatives.
 If the reference directory has been enhanced via the above-described system of metatagging, then the match can be undertaken via sophisticated statistical matching techniques. This would be the case in a closed network, like a cable television system, where the system operator controlled all alternatives.
 If the reference directory is maintained without such enhancement, but is capable of responding to Boolean ANDed queries (as would be the case with an open network like the Internet, where no individual or institution controls the array of alternatives, but search engines provide a directory for access to all alternatives) then the matching can be accomplished through the real-time metatagging and automated submission of Boolean ANDed queries derived from the metatagging process, as described in our “Brilliant Query” application.
 There will be cases where the number of recommended alternatives displayed on a given screen needs to be limited. When the number of alternatives matching the keywords for a given screen exceeds this limit, there are one or more criteria for determining the alternatives that should be displayed. First, all alternatives that are associated with all keywords are selected as a set. One or more of the following optional criteria then orders the recommendation list:
 1) Date of listing (how new or time-sensitive is the content, product or service).
 2) Keyword weighting (a content, product or service that matches the keyword that most frequently appears in the source article from the directory will be displayed before a content, product or service that matches a lower frequency keyword).
 The list of suggested products and/or services is then limited to a system operator-specified number of alternatives based on the screen design preferences.
 Metatagged keywords are simply a collection of words, generated automatically or manually, that are deemed to be indicative of the topic matter or one of the topics for a given directory selection. Metatagged keywords are determined by comparison of a pre-determined list of keywords to the text of the directory selection. If the directory selection contains one or more of the keywords, or an appropriate synonym, then that keyword is associated with that text body and potentially used for the metatag-enhanced datamining. Keywords may also be determined by statistical word frequency analysis of the text, with or without manual selection and addition of synonyms.
 Keywords are generated by automatic or manual statistical and empirical analysis of the body of content to be enhanced or a comparable body of content. The keyword list for a given content source is generated through the use of word frequency analysis, stopword removal and finally, manual selection using empirical testing of the results generated by a given potential keyword. Based on experience, a solid keyword list usually runs between 250 and 1000 words and phrases, which are chosen by the system designer. The keyword set can be periodically improved and optimized by word frequency analysis of large bodies of text, which represent the appropriate territories of interest.
 Also, the keyword set can be manually tuned through the use of a thesaurus feature whereby a given keyword can be associated with one or more synonyms that would indicate the use of the keyword whenever one or more of the synonyms appear in the body of text to be enhanced.
 Marketers use datamining to identify which clusters of consumers perform at above or below average rates for their content, products and services, by responding to their offers. By identifying the consumers that are navigating to specific content, products and/or services on a network by their underlying interests, marketers will be able to prove they can have a “care for me” relationship and a “you know who I am” relationship with the consumer.
 The statistical analysis of user navigation data fortified with metatagged keywords, whether conducted locally in a specific ‘client’ device or at a global collection point for all users (in the datacube) will identify frequently occurring patterns of behavior.
 By definition, high frequency behavior means valuable clusters of consumers. Once identified, system operators will act in three territories to exploit the traffic analysis.
 1) GUI: Knowing what people are interested in will allow system operators to redesign their graphic user interfaces (GUI) to put the content, products and services most likely to be of interest ‘up front’, the most important news stories at the top of the page, the most interesting television shows at the beginning of the program guide, the most attractive merchandising or promotional offers in the context of the video-on-demand offering of greatest interest to the related audience cluster. All these are examples of the optimization of GUI in response to the results of the metatag-driven traffic analysis.
 2) Recommendation Engine: Where a vast array of alternative selections are available, metatag-driven data analysis will allow similar content, products and services to be recommended to consumer clusters.
 3) Value-based sales: System operators, equipped with empirical evidence of what content, products and services around which consumers of common interests are clustering, will be able to value their advertising space and price their sponsorships on the basis of constantly sensed consumer behavior data. Thorough and constant traffic analysis on a global basis will yield highly precise descriptions of commonly recurring consumer behavior patterns. These patterns will function like a well-researched map, and when compared to the less robust data analysis (limited because the local data is limited) performed in local (client) devices, like PCs and set-top boxes, will allow the recognition of less-well defined local behavior patterns, thus enabling ‘best guess’ GUI adjustments, recommendations and targeted advertising and sponsorships.
FIG. 1 is a depiction of a news story embellished with metatagged keywords;
FIG. 2 is a depiction of a computer screen showing a metatagged news story, for which ‘Brilliant Queries” and “Targeted e-commerce” offers have been created, by exploitation of the keyword metatagging;
FIG. 3 shows examples of five different HTML commands which are embedded in the news story web page of FIG. 2, in order to enable the signaling of consumer navigation behavior ‘upstream’ to a datamining system;
FIG. 4 is a depiction of a computer screen showing a spot report on searches made in Encyclopedia Electronica;
FIG. 5 is a depiction of a computer screen showing a real-time Internet delivered daily report;
FIG. 6 is a sample of a printed summary report based on traffic analysis data;
FIG. 7 is a depiction of a computer screen showing a Graphic User Interface (GUI); and,
FIG. 8 is a depiction of a computer screen showing an e-commerce offer, optimized and targeted on the basis of traffic analysis data.
FIG. 1 is a news story entitled “Lazio Accuses Clinton”. embellished with metatagged keywords associated with the article. The keywords have been chosen by the use of word frequency analysis and keyword lists, as described in our parent patent applications.
FIG. 2 is a depiction of the same metatagged news story. for which “Brilliant Queries” and “Targeted e-commerce”, offers have been created by exploitation of the keyword metatagging.
FIG. 3 shows examples of five different HTML commands which would be embedded in the news story web page of FIG. 2, in order to enable the signaling of consumer navigation behavior ‘upstream’ to a datamining system which tracks consumer data.
FIG. 4 depicts a spot report on searches made in Encyclopedia Electronica, where every article has been metatagged, and consumer search behavior is cross-correlated to the metatagged keywords associated with every article, hence revealing statistically significant consumer interests through traffic analysis.
FIG. 5 depicts a real-time internet-delivered daily report based on data derived from datamining data derived from Encyclopedia Electronica, exposing the frequency of occurrence of metatagged keywords.
FIG. 6 shows a sample of a printed summary report based on traffic analysis data. Shown are the rankings of areas of interest listed in order of pages having been viewed by users (at the latest time), with the two previous times also disclosed. Also shown are the top related stories for each area of interest.
FIG. 7 depicts a screen page of a Graphic User Interface (GUI), which is constantly optimized on the basis of traffic analysis data. The ‘Top Ten’ searches, and the ‘Top Ten’ articles are chosen and sorted on the basis of traffic analysis data, i.e. those stories and those articles, which prove to be of the greatest interest to consumers.
FIG. 8 depicts a screen page of an e-commerce offer, optimized and targeted on the basis of traffic analysis data. The keywords are chosen and sorted on the basis of traffic analysis, and the offer delivered to only those consumers whose interests match the ‘keyword profile’ of the merchandise offer.
 Having thus described the invention.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US2151733||May 4, 1936||Mar 28, 1939||American Box Board Co||Container|
|CH283612A *||Title not available|
|FR1392029A *||Title not available|
|FR2166276A1 *||Title not available|
|GB533718A||Title not available|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US6877002 *||Jun 11, 2001||Apr 5, 2005||America Online, Inc.||Fuzzy database retrieval|
|US7043555 *||Jun 30, 2004||May 9, 2006||Novell, Inc.||Techniques for content filtering|
|US7177608||Mar 10, 2003||Feb 13, 2007||Catch A Wave Technologies||Personal spectrum recorder|
|US7464086||Sep 14, 2001||Dec 9, 2008||Yahoo! Inc.||Metatag-based datamining|
|US7506055||Jun 28, 2004||Mar 17, 2009||Novell, Inc.||System and method for filtering of web-based content stored on a proxy cache server|
|US7617176 *||Jul 13, 2004||Nov 10, 2009||Microsoft Corporation||Query-based snippet clustering for search result grouping|
|US7702541||Dec 15, 2000||Apr 20, 2010||Yahoo! Inc.||Targeted e-commerce system|
|US7707208||Oct 10, 2006||Apr 27, 2010||Microsoft Corporation||Identifying sight for a location|
|US7720836||Jul 7, 2004||May 18, 2010||Aol Inc.||Internet streaming media workflow architecture|
|US7752186||Dec 20, 2004||Jul 6, 2010||Aol Inc.||Grouping multimedia and streaming media search results|
|US7836050||Jan 25, 2006||Nov 16, 2010||Microsoft Corporation||Ranking content based on relevance and quality|
|US7925967||Jun 8, 2001||Apr 12, 2011||Aol Inc.||Metadata quality improvement|
|US8060451 *||Jun 15, 2007||Nov 15, 2011||International Business Machines Corporation||System and method for facilitating skill gap analysis and remediation based on tag analytics|
|US8095529||Jan 4, 2005||Jan 10, 2012||Aol Inc.||Full-text relevancy ranking|
|US8151291||Jun 11, 2007||Apr 3, 2012||The Nielsen Company (Us), Llc||Methods and apparatus to meter content exposure using closed caption information|
|US8209311||Jul 1, 2010||Jun 26, 2012||Aol Inc.||Methods and systems for grouping uniform resource locators based on masks|
|US8515937 *||Jun 29, 2009||Aug 20, 2013||Alexa Internet||Automated identification and assessment of keywords capable of driving traffic to particular sites|
|US8527520||Jan 11, 2012||Sep 3, 2013||Streamsage, Inc.||Method and system for indexing and searching timed media information based upon relevant intervals|
|US8533223||May 12, 2009||Sep 10, 2013||Comcast Interactive Media, LLC.||Disambiguation and tagging of entities|
|US8589328 *||Mar 31, 2009||Nov 19, 2013||Symantec Corporation||Method and apparatus for examining computer user activity to assess user psychology|
|US8595475||Feb 17, 2012||Nov 26, 2013||AOL, Inc.||Method of disseminating advertisements using an embedded media player page|
|US8700590||Jun 20, 2012||Apr 15, 2014||Microsoft Corporation||Grouping multimedia and streaming media search results|
|US8706735 *||Jul 31, 2013||Apr 22, 2014||Streamsage, Inc.||Method and system for indexing and searching timed media information based upon relevance intervals|
|US8713016||Dec 24, 2008||Apr 29, 2014||Comcast Interactive Media, Llc||Method and apparatus for organizing segments of media assets and determining relevance of segments to a query|
|US8738486 *||Dec 31, 2007||May 27, 2014||Mastercard International Incorporated||Methods and apparatus for implementing an ensemble merchant prediction system|
|US8819404||Oct 9, 2013||Aug 26, 2014||Aol Inc.||Method of disseminating advertisements using an embedded media player page|
|US8918812||Oct 24, 2001||Dec 23, 2014||Aol Inc.||Method of sizing an embedded media player page|
|US9009136||May 14, 2010||Apr 14, 2015||Microsoft Technology Licensing, Llc||Methods and systems for enhancing metadata|
|US9110931||Feb 8, 2005||Aug 18, 2015||Microsoft Technology Licensing, Llc||Fuzzy database retrieval|
|US20020016741 *||Dec 15, 2000||Feb 7, 2002||Black Peter M.||Targeted e-commerce system|
|US20040045040 *||Oct 24, 2001||Mar 4, 2004||Hayward Monte Duane||Method of sizing an embedded media player page|
|US20040064500 *||Nov 20, 2001||Apr 1, 2004||Kolar Jennifer Lynn||System and method for unified extraction of media objects|
|US20040193589 *||Feb 11, 2004||Sep 30, 2004||Hitachi Software Engineering Co., Ltd.||Key word frequency calculation method and program for carrying out the same|
|US20050021796 *||Jun 28, 2004||Jan 27, 2005||Novell, Inc.||System and method for filtering of web-based content stored on a proxy cache server|
|US20050177568 *||Jan 4, 2005||Aug 11, 2005||Diamond Theodore G.||Full-text relevancy ranking|
|US20060026152 *||Jul 13, 2004||Feb 2, 2006||Microsoft Corporation||Query-based snippet clustering for search result grouping|
|US20060212441 *||Oct 25, 2005||Sep 21, 2006||Yuanhua Tang||Full text query and search systems and methods of use|
|US20080153520 *||Dec 21, 2006||Jun 26, 2008||Yahoo! Inc.||Targeted short messaging service advertisements|
|US20090171759 *||Dec 31, 2007||Jul 2, 2009||Mcgeehan Thomas||Methods and apparatus for implementing an ensemble merchant prediction system|
|US20130318121 *||Jul 31, 2013||Nov 28, 2013||Streamsage, Inc.||Method and System for Indexing and Searching Timed Media Information Based Upon Relevance Intervals|
|EP1462954A2 *||Feb 10, 2004||Sep 29, 2004||Hitachi Software Engineering Co., Ltd.||Key word frequency calculation method and program for carrying out the same|
|EP2030439A2 *||Jun 11, 2007||Mar 4, 2009||The Nielsen Company||Methods and apparatus to meter content exposure using closed caption information|
|WO2006047654A2 *||Oct 25, 2005||May 4, 2006||Qianjin Hu||Full text query and search systems and methods of use|
|U.S. Classification||1/1, 707/999.104, 707/999.001|
|Cooperative Classification||G06Q30/02, Y10S707/99935|
|Sep 14, 2001||AS||Assignment|
Owner name: ETRONICA.COM CORPORATION, CALIFORNIA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:BLACK, PETER M.;WATERS, ANTHONY BRYAN;REEL/FRAME:012174/0960
Effective date: 20010912
|Aug 15, 2005||AS||Assignment|
Owner name: OVERTURE SERVICES, INC.,CALIFORNIA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:ETRONICA, INC.;REEL/FRAME:016883/0756
Effective date: 20050718
|Oct 9, 2008||AS||Assignment|
Owner name: YAHOO! INC,CALIFORNIA
Free format text: MERGER;ASSIGNOR:OVERTURE SERVICES, INC;REEL/FRAME:021652/0654
Effective date: 20081001
|May 9, 2012||FPAY||Fee payment|
Year of fee payment: 4